<ul><li>A study investigates the practice of discarding unimportant features based on small aggregate SHAP values.</li><li>The study finds that small aggregate SHAP values do not necessarily imply that the corresponding feature has no effect on the function.</li><li>To address this issue, the study suggests aggregating SHAP values over the extended support, which is the product of the marginals of the underlying distribution.</li><li>The study also extends the findings to KernelSHAP, demonstrating that a small aggregate value justifies feature removal, regardless of the accuracy of the KernelSHAP approximation.</li></ul>

How to safely discard features based on aggregate SHAP values

Discover more